And OpenClaw is nothing revolutionary. It’s all shit we could do before OpenClaw. It’s just that no one was stupid enough to do it. Now everyone has gone crazy.
but isn't this like a lot of other CS-related "gradient descent"?
when someone invents a new scheduling algorithm or a new concurrent data structure, it's usually based on hunches and empirical results (benchmarks) too. nobody sits down and mathematically proves their new linux scheduler is optimal before shipping it. they test it against representative workloads and see if there is uplift.
we understand transformer architectures at the same theoretical level we understand most complex systems. we know the principles, we have solid intuitions about why certain things work, but the emergent behavior of any sufficiently complex system isn't fully predictable from first principles.
that's true of operating systems, distributed databases, and most software above a certain complexity threshold.
No. Algorithm analysis is much more sophisticated and well defined than that. Most algorithms are deterministic, and it is relatively straightforward to identify complexity, O(). Even nondeterministic algorithms we can evaluate asymptotic performance under different categories of input. We know a lot about how an algorithm will perform under a wide variety of input distributions regardless of determinism. In the case of schedulers, and other critical concurrency algorithms, performance is well known before release. There is a whole subfield of computer science dedicated to it. You don't have to "prove optimality" to know a lot about how an algorithm will perform. What's missing in neural networks is the why and how any inputs will propagate, through the network during inference. It is a black box of understandability. Under a great deal of study, but still very poorly understood.
i agree w/ the the complexity analysis point, but that theoretical understanding actually translates to real world deployment decisions in both subfields. knowing an algorithm is O() tells you surprisingly little about whether itll actually outperform alternatives on real hardware with real cache hierarchies, branch predictors, and memory access patterns. same thing with ML (just with the very different nature of GPU hw), both subfields hve massive graveyards of "improvements" that looked great on paper (or in controlled environments) but never made it into production systems. arxiv is full of architecture tweaks showing SOTA on some benchmark and the same w/ novels data structures/algorithms that nobody ever uses at scale.
I think you missed the point. Proving something is optimal, is a much higher bar than just knowing how the hell the algorithm gets from inputs to outputs in a reasonable way. Even concurrent systems and algorithm bounds under input distributions have well established ways to evaluate them. There is literally no theoretical framework for how a neural network churns out answers from inputs, other than the most fundamental "matrix algebra". Big O, Theta, Omega, and asymptotic performance are all sound theoretical methods to evaluate algorithms. We don't have anything even that good for neural networks.
yeah, but isn't the whole point of claude code to get people to provide preference data/telemetry data to anthropic (unless you opt out?). same w/ other providers.
i'm guessing most of the gains we've seen recently are post training rather than pretraining.
Yes, but you have the problem that a good portion of that is going to be AI generated.
But, I naively assume most orgs would opt out. I know some orgs have a proxy in place that will prevent certain proprietary code from passing through!
This makes me curious if, in the allow case, Anthropic is recording generated output, to maybe down-weight it if it's seen in the training data (or something similar)?
yup, bezos said "we will be able to beat the cost of terrestrial data centers in space in the next couple of decades". presumably this means they'll need huge ass radiators, so its all about bringing down launch costs since they'll need to increase mass.
they should just acquire one of the many agent code harnesses. Something like opencode works just as well as claude-code and has only been around half of the time.
I used opencode happily for a while before switching to copilot cli. Been a minute , but I don't detect a major quality difference since they added Plan mode. Seems pretty solid, and first party if that matters to your org.
true on the naming, but i think geometric/clifford algebra has its own mysterious aura precisely because it can be framed as "suppressed" or "overlooked".. plus it genuinely does have elegant mathematical structure backing up the hype
funny thing is quaternions had that exact same energy in the computer graphics community for years. after ken shoemake introduced them to CG in 1985, there was a long period of "why are we using euler angles like cavemen when this exists??". now quaternions are well known tooling for people in graphics and the mystique has worn off at least in that community.
>EDIT: More interestingly, I find an issue, what do I even DO? If it's not related to integrations or your underlying data, the black box just gave nonsensical output. What would I do to resolve it?
Lots of stuff you could do. Adjust the system prompt, add guardrails/filters (catching mistakes and then asking the LLM loop again), improve the RAG (assuming they have one), fine tune (if necessary), etc.
reply